DeepSeek previews new architecture: 552B MoE with just 8B active input parameters
zephyr_z9 · x · 2026-09-10
DeepSeek officially previewed its next-gen model: an asymmetric 'more intelligence, less cost' Causal Encoder–Decoder MoE with 552B total parameters but only 8B active for input and 16B for output. New pre-training methods plus larger-scale RL post-training reportedly deliver benchmark results ahead of flagship models including DeepSeek-V4-Pro. Details to follow in a 6-part thread.
Related event: DeepSeek Unveils Open-Source V4.1-Flash MoE Model(28 posts)→
More from Models
- DeepSeek V4.1 Flash is actually 748B params, safetensors analysis shows — DistanceSolar1449 · 2026-09-10
- Kimi K3 lands on RunPod: 2.8T params, 1M context, $3/$15 per 1M tokens — Kimi_Moonshot · 2026-09-10
- Dev burns 300M tokens on GLM 5.3 in a week and still has quota left — saibharadwaj · 2026-09-10
- Follow-up: a 3T-parameter model may already exist, scaling issues remain the wildcard — teortaxesTex · 2026-09-10
- Speculation: DeepSeek V4.1 Pro could be a 3.1T-param MoE with 2.6TB disk footprint — teortaxesTex · 2026-09-10
- New Book Teaches Beginners to Build and Fine-Tune Their Own GPT-Style SLMs, With Colab Notebooks — Roger_M_Taylor · 2026-09-10